On this page

Run Command

cyberstrike run executes a single prompt non-interactively — ideal for automation and scripting. In this mode there is nobody to answer permission prompts, so any tool that would prompt is auto-rejected (see Automation).

📸 SCREENSHOT: run-command-output.png

Run command execution output

Basic Usage

cyberstrike run "your prompt here"

The prompt is also read from stdin when the input is piped:

echo "summarize the findings in report.json" | cyberstrike run

Options

Flag	Alias	Description
`--model`	`-m`	Model to use, as `provider/model`
`--agent`		Primary agent to use (only `cyberstrike` ships built-in)
`--format`		Output format: `default` or `json` (default: `default`)
`--file`	`-f`	File(s) to attach to the message (repeatable)
`--continue`	`-c`	Continue the last session
`--session`	`-s`	Session id to continue
`--fork`		Fork the session before continuing (needs `--continue`/`--session`)
`--share`		Share the session
`--title`		Title for the session
`--variant`		Model variant / reasoning effort (e.g. `high`, `max`, `minimal`)
`--thinking`		Show thinking blocks
`--attach`		Attach to a running server (e.g. `http://localhost:4096`)
`--dir`		Directory to run in (remote path when attaching)
`--port`		Port for the local server
`--command`		Run a named command; use the message for its arguments

Caution

There is no --output, --permission, --timeout, --verbose, or --quiet flag. Output control is --format (default | json); reasoning display is --thinking.

Examples

JSON output for scripting

cyberstrike run "scan example.com for OWASP Top 10 issues" --format json > report.json

Pick a model and attach files

cyberstrike run "review this code for auth bugs" \
  -m anthropic/claude-sonnet-4-20250514 \
  -f src/auth.ts -f src/session.ts

Continue or fork a session

cyberstrike run "now test the admin endpoints" --continue
cyberstrike run "try a different payload" --session <id> --fork

Permissions in `run`

Because prompts can’t be answered non-interactively, pre-approve the tools a script needs with the CYBERSTRIKE_PERMISSION env var or the config permission field:

CYBERSTRIKE_PERMISSION='{"bash":"allow","read":"allow"}' \
  cyberstrike run "enumerate services on the host" --format json

See Automation for CI patterns.

Exit Codes

cyberstrike run returns 0 on success and a non-zero code on error (e.g. invalid arguments) or interruption.

CLI Reference - All commands
Automation - CI/CD and pre-approved permissions
Usage: CLI - Non-interactive usage